We discuss the interferometric scheme put forward by Tan, Walls, and Collett (TWC) that claims to reveal Bell non-classicality of a single photon. We show that the violation of local realism they report hinges on an additional assumption about hidden variables. Without it, one can construct a classical model that correctly predicts all the probabilities of the TWC scheme. To illustrate this, we construct a proper Bell inequality, devoid of additional assumptions, that is not violated by the TWC setup. We then show a modification of the interferometric scheme, consisting in making the amplitude of auxiliary fields tunable, that enables the Bell violation. Finally, we present an alternative point of view on the shortcomings of the original TWC analysis, arguing that it effectively treats the weak coherent states classically.